DDRNet23-Slim: Optimized for Qualcomm Devices

DDRNet23Slim is a machine learning model that segments an image into semantic classes, specifically designed for road-based scenes. It is designed for the application of self-driving cars.

This is based on the implementation of DDRNet23-Slim found here. This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the Qualcomm® AI Hub Models library to export with custom configurations. More details on model performance across various devices, can be found here.

Qualcomm AI Hub Models uses Qualcomm AI Hub Workbench to compile, profile, and evaluate this model. Sign up to run these models on a hosted Qualcomm® device.

Getting Started

There are two ways to deploy this model on your device:

Option 1: Download Pre-Exported Models

Below are pre-exported model assets ready for deployment.

Runtime	Precision	Chipset	SDK Versions	Download
ONNX	float	Universal	QAIRT 2.42, ONNX Runtime 1.25.0	Download
ONNX	w8a8	Universal	QAIRT 2.42, ONNX Runtime 1.25.0	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download
TFLITE	w8a8	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit DDRNet23-Slim on Qualcomm® AI Hub.

Option 2: Export with Custom Configurations

Use the Qualcomm® AI Hub Models Python library to compile and export the model with your own:

Custom weights (e.g., fine-tuned checkpoints)
Custom input shapes
Target device and runtime configurations

This option is ideal if you need to customize the model beyond the default configuration provided here.

See our repository for DDRNet23-Slim on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.semantic_segmentation

Model Stats:

Model checkpoint: DDRNet23s_imagenet.pth
Inference latency: RealTime
Input resolution: 2048x1024
Number of output classes: 19
Number of parameters: 6.13M
Model size (float): 21.7 MB
Model size (w8a8): 6.11 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
DDRNet23-Slim	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	10.414 ms	29 - 258 MB	NPU
DDRNet23-Slim	ONNX	float	Snapdragon® X2 Elite	10.853 ms	188 - 188 MB	NPU
DDRNet23-Slim	ONNX	float	Snapdragon® X Elite	27.58 ms	157 - 157 MB	NPU
DDRNet23-Slim	ONNX	float	Snapdragon® 8 Gen 3 Mobile	19.281 ms	32 - 310 MB	NPU
DDRNet23-Slim	ONNX	float	Qualcomm® QCS8550 (Proxy)	28.253 ms	0 - 21 MB	NPU
DDRNet23-Slim	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	13.571 ms	6 - 203 MB	NPU
DDRNet23-Slim	ONNX	float	Qualcomm® QCS9075	39.221 ms	24 - 93 MB	NPU
DDRNet23-Slim	ONNX	float	Qualcomm® QCS8750	13.571 ms	6 - 203 MB	NPU
DDRNet23-Slim	ONNX	float	Qualcomm® QCS7181	27.58 ms	157 - 157 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	39.528 ms	7 - 206 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® X2 Elite	39.667 ms	206 - 206 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® X Elite	55.838 ms	175 - 175 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	39.787 ms	7 - 254 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® QCS6490	299.947 ms	201 - 218 MB	CPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	53.68 ms	4 - 19 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	250.094 ms	146 - 155 MB	CPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® QCM6690	267.22 ms	210 - 219 MB	CPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	37.81 ms	1 - 204 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® QCS9075	52.918 ms	6 - 51 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® QCS7790	250.094 ms	146 - 155 MB	CPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® QCS8750	37.81 ms	1 - 204 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® QCS7181	55.838 ms	175 - 175 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	10.428 ms	10 - 244 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Snapdragon® X2 Elite	11.842 ms	24 - 24 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Snapdragon® X Elite	34.088 ms	24 - 24 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	22.336 ms	21 - 290 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® QCS8275	98.398 ms	15 - 212 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	33.033 ms	24 - 27 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® SA8775P	40.43 ms	24 - 221 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® SA8650P	40.43 ms	24 - 221 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® SA8255P	40.43 ms	24 - 221 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	66.519 ms	5 - 282 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® SA7255P	98.398 ms	15 - 212 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® SA8295P	43.541 ms	24 - 229 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	15.511 ms	17 - 237 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® QCS9075	53.896 ms	24 - 52 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® QCS8750	15.511 ms	17 - 237 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® QCS7181	34.088 ms	24 - 24 MB	NPU
DDRNet23-Slim	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	10.362 ms	0 - 237 MB	NPU
DDRNet23-Slim	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	22.313 ms	1 - 282 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® QCS8275	98.343 ms	2 - 204 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® QCS8550 (Proxy)	33.142 ms	2 - 37 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® SA8775P	40.415 ms	3 - 205 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® SA8650P	40.415 ms	3 - 205 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® SA8255P	40.415 ms	3 - 205 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® QCS8450 (Proxy)	66.623 ms	0 - 284 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® SA7255P	98.343 ms	2 - 204 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® SA8295P	43.413 ms	2 - 216 MB	NPU
DDRNet23-Slim	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	15.38 ms	2 - 230 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® QCS9075	53.397 ms	0 - 41 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® QCS8750	15.38 ms	2 - 230 MB	NPU

License

The license for the original implementation of DDRNet23-Slim can be found here.

References

Community

Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
For questions or feedback please reach out to us.

Downloads last month: -; Downloads are not tracked for this model. How to track

Paper for qualcomm/DDRNet23-Slim

Deep Dual-resolution Networks for Real-time and Accurate Semantic Segmentation of Road Scenes

Paper • 2101.06085 • Published Jan 15, 2021